Consensus Hack: US Politicians Use GEO to Manipulate ChatGPT Responses

The information integrity of large language models has encountered a novel, systemic threat vector. On October 26, 2026, Reuters published an investigation exposing the deployment of GEO (Generative Engine Optimization) by US political campaigns. The campaign of Republican Mike Minogue systematically generated and seeded thousands of political artifacts across the web, specifically engineered to be ingested by the RAG (Retrieval-Augmented Generation) pipelines powering ChatGPT, Claude, and Gemini during voter queries.

GEO represents the weaponized evolution of SEO, architected to hack the retrieval and training loops of AI systems. The objective is to manipulate models into synthesizing biased output and issuing targeted political endorsements. For media macroeconomics, this shatters the illusion of algorithmic objectivity. Chatbots have emerged as highly vulnerable targets for coordinated information warfare. With millions of users placing higher implicit trust in LLM summaries than traditional Google search results, manipulating the semantic "memory" of these agents has become the primary weapon for political technologists, forcing frontier labs to urgently overhaul source-attribution architectures.

Источники: Reuters